PRESTACIONES BÁSICAS Y COMPLEMENTARIAS

Protein-protein interactions play crucial roles in regulating important cellular pathways. Therefore, novel potential interacting partners of SOX2 were investigated using yeast-2- hybrid (y2h) pairwise experiments and bimolecular fluorescence complementation (BiFC) experiments in collaboration with the DKFZ proteomics core facility (methods described in Supplemental section 8.1.2.2 and 8.1.2.3)

Using BiFC, the potential interaction between SOX2 and PAX3, SOX10, and PAX6, respectively, was investigated. PAX6 is a known interacting partner of SOX2, particularly in the retina and was used as a positive control [245]. HEK 293T cells were transfected with bait and prey vectors containing fragments of complementary fluorescent reporter proteins. Upon interaction in vitro, the fusion reporter protein reforms and fluoresces. A positive interaction between PAX3 and SOX2 was observed in three independent BiFC experiments, similar to that of the positive control (Figure 14a, left panel). The differential role of SOX2 and SOX10 was investigated in melanoma progression and therapy and it was found that SOX2 and SOX10 share a binding site on the proximal MITF promoter. Therefore, the mechanism, by which SOX2 and SOX10 may be regulated on protein-level, was of interest. Strong fluorescence was observedwhen SOX2-bait and SOX10-prey were transfected into HEK 293T cells, where the fluorescence was equal to positive control (Figure 14a, middle panel). The same fluorescence was observed when SOX10 was used as bait and SOX2 as prey (Figure 14a, middle panel).

Next, to confirm these interactions in a second independent experiment, the y2h pairwise experiment was performed. In brief, a construct containing the bait (SOX2) coupled to an enzyme capable of synthesizing tryptophan (TRP) and the prey of our choice coupled to an enzyme capable of synthesizing leucine (LEU) were transformed into yeast. The yeasts were then mated and diploid yeast were selected via nutritional selection on TRP/LEU deficient medium (Figure 14b). Growth of the diploid yeast was only possible upon interaction of the bait, containing the protein binding domain, and the prey, containing the activation domain (Figure 14c). This interaction allows for the reformation of a transcription factor which initiates transcription of the integrated HIS3 gene, enabling growth on histidine-deficient media. Therefore, only diploid yeast clones in which there is an interaction between the bait and prey proteins are able to transcribe their own HIS3 and survive on histidine-deficient medium. With this selection process, one is able to elucidate potential protein-protein interactions. Another level of stringency used in y2h experiments is 3-amino-1,2,4-triazole (3- AT). This molecule is a competitive inhibitor of the product of the HIS3 gene, reducing possible autoactivation and false positives.

The same interactions, PAX3/SOX2 and SOX10/SOX2, identified by the BiFC assay were further examined in y2h experiments (Figure 14d). However, difficulties were observed when performing these experiments due to the strong autoactivation tendencies of SOX2. Therefore, the effect of autoactivation could not be completely ruled out in y2h pairwise experiments. Nonetheless, weak colony growth in the PAX3/SOX2 mating at the highest stringency of 2.0 mM 3-AT (bottom panels) was observed; however between replicates (column) there was great variability (Figure 14d, left panel). Next, the SOX2/SOX10 potential interaction was investigated in the y2h system; however the overexpression of both SOX2 and SOX10 proved to be lethal in yeast even in the lowest stringency tested (Figure 14d, upper panels). Therefore, the possible interaction between SOX2 and SOX10 was not confirmed in yeast (Figure 14d, right panel).

TWIST1 was further investigated as a potential interactor with SOX2. TWIST1 is known to be implicated in cell lineage determination and the EMT process. Since SOX2 plays important roles in both pathways possible protein-protein interactions between these two proteins were investigated. Unfortunately, TWIST1 analyses could not be performed using BiFC experiments due to cloning difficulties. Nonetheless, y2h pairwise experiments for TWIST1 were performed. Although autoactivation-mediated effects could not be completely ruled out in all y2h pairwise experiments, weak colony growth in the TWIST1-SOX2 mating was observed at the highest stringency of 2.0 mM 3-AT (bottom panels); however between replicates (columns) there was great variability (Figure 14d, middle panels). In conclusion, SOX2 may form potential protein-protein interactions with PAX3, SOX10 and TWIST1; however since the y2h pairwise experiments were unreliable due to strong autoactivation of SOX2 and therefore these experiments need to be repeated with SOX2 construct lacking the HMG domain to eliminate autoactivation tendencies. Furthermore, these interactions should be confirmed in co-immunoprecipitation experiments in vitro.

5.6.1 SOX2 forms protein-protein interactions with novel candidates related to

SUMOylation

In addition to y2h pairwise experiments, y2h high-throughput screen (HTS) was performed. We performed the screen on multiple libraries (human universal, full ORF N-terminal, full ORF C-terminal) to eliminate as many artifacts as possible. The concentration of 3-AT used depends on the strength of the interaction between the bait and prey used and is usually determined in preliminary experiments (Figure S5a). During this HTS, 3-AT was used at a concentration of 4 mM to eliminate as many promiscuous and artificial preys. The experimental setup of the screen is identical to the basics of y2h experiments, as described in Figure 14b-c. The main difference between the standard y2h pairwise experiments and the HTS is that in the screen, the prey proteins are not pre-selected but contained in a cDNA

library. Therefore, the mating occurred between a particular bait of interest (SOX2) and the prey offered by the cDNA library. Upon the reformation of the transcription factor, reporter genes were transcribed allowing fluorescence readout. After mating, the yeast were dispensed onto ten microtiter plates and allowed to grow (Figure S5b). Fluorescence was measured and the positive wells collected. Next, PCR was performed and the bands were sequenced and analyzed.

Three independent HTSs were performed using three different cDNA libraries: universal human, full ORF library (N-terminal) and full ORF library (C-terminal). Several hits were isolated in at least two of the three screens. These hits included SUMO1, UBE2I, PIAS4, PCFG2, and CBX4 (Figure 14e). For example, small ubiquitin-related modifier 1 (SUMO1) was isolated 14 times from the universal human library and 37 times from the full ORF library. Ubiquitin-conjugating enzyme E2I (UBE2I) was isolated 11 times from the universal human library and 29 times from the full ORF library (C-terminal) (Figure 14e).

The first hit, SUMO1, was particularly interesting since this protein is heavily involved in protein SUMOylation or post-translational modifications [252]. Literature research revealed that the conjugation of SOX2 to SUMO1 has already been investigated [252]. Moreover, it was discovered that upon conjugation of SUMO1 to SOX2, inhibition of the DNA binding activity by SOX2 was observed. This work not only suggests that SUMOylation has a functional effect on SOX2 but also validates that the y2h HTS likely identified true protein- protein interactions. Likewise, the next hit, ubiquitin-conjugating enzyme (UBE) - 2I, is an enzyme responsible for performing SUMO conjugation is related to the SUMOylation process. Further hits, E3 SUMO-protein ligase (PIAS4) and Chromobox protein homolog 4 (CBX4), function in the SUMOylation process by stabilizing the interaction between UBE2I and the substrate [247]. The identification of potential interactions occurring between SOX2 and proteins related to the SUMOylation process suggest SOX2 can be SUMOylated.

To determine whether SOX2 can be SUMOylated, the GPS-SUMO online database was used to predict SUMOlyation sites within the SOX2 protein (Figure 14f). The database identified two predicted SUMOylation sites for SOX2; one located at amino acid position 121 and the other at 245. Interestingly, the publication which reported conjugation of SOX2 to SUMO1 describes a conjugation at lysine 247 [252]; but the exact positions where SOX2 is SUMOylated in vitro needs to be confirmed. Lastly, the SOX2 protein domains and its predicted SUMOylation sites were constructed using the open source software DOG 2.0 (Figure 14g).

Taken together, y2h HTS experiments identified four key players in the SUMOylation process, SUMO1, UBE2I, PIAS4 and CBX4, which may form protein-protein interactions with SOX2.